Search papers, labs, and topics across Lattice.
HiFlow is introduced as a hierarchical feedback-driven optimization framework to improve constrained long-form text generation by addressing the limitations of static planning and offline supervision. It decomposes generation into a planning layer for global structure and a generation layer for conditioned text, enabling joint optimization. Experiments across various backbones demonstrate HiFlow's superior performance in generating high-quality, constraint-satisfying long texts compared to existing baselines.
Achieve better long-form text generation under constraints by jointly optimizing global planning and local generation with hierarchical feedback.
Large language models perform well in short text generation but still struggle with long text generation, particularly under complex constraints. Such tasks involve multiple tightly coupled objectives, including global structural consistency, local semantic coherence, and constraint feasibility, forming a challenging constrained optimization problem. Existing approaches mainly rely on static planning or offline supervision, limiting effective coordination between global and local objectives during generation. To address these challenges, we propose HiFlow, a hierarchical feedback-driven optimization framework for constrained long text generation. HiFlow formulates generation as a two-level optimization process, consisting of a planning layer for global structure and constraint modeling, and a generation layer for conditioned text generation. By incorporating constraint-aware plan screening and closed-loop feedback at both levels, HiFlow enables joint optimization of planning quality and generation behavior, progressively guiding the model toward high-quality, constraint-satisfying outputs. Experiments on multiple backbones confirm HiFlow's effectiveness over baseline methods.
